Kay Hayen comments

Results 1287 comments of


                                            Kay Hayen

Wrong optimization with nested loops where only the inner loop changes a variable

You are not on the latest hotfix, but even 1.5 shows this. Of course has nothing to do with standalone code, the given example does it in accelerated mode. What's...

Wrong optimization with nested loops where only the inner loop changes a variable

This is a bug of the loop shape analysis, I believe. It seems that the operation is considered between ` `, and then suddenly ` `. What is going on...

Wrong optimization with nested loops where only the inner loop changes a variable

Ok, the `type` optimization looks at the type shape, and only if it has a type name, resolves it to a built-in name. That doesn't cover all the types, esp....

Wrong optimization with nested loops where only the inner loop changes a variable

The reproducer is with that new optimization added down to this one: ``` def __build_word_numeral(*args): re_ = None for word_list in args: for word in word_list: if not re_: re_...

Wrong optimization with nested loops where only the inner loop changes a variable

So the nested loop analysis is going wrong. I believe this must be because the outer loop concludes that nothing is going on with `re_` and therefore the value is...

Wrong optimization with nested loops where only the inner loop changes a variable

I will try and aim at a loop analysis rewrite for 1.7 or 1.8, it's not going to fly this way for all cases.

Wrong optimization with nested loops where only the inner loop changes a variable

This rewrite is on for 2.0 I believe, but it's not clear how much effort it is. Mainly it's about scalability, making the merge traces static operations, and the hope...

Wrong optimization with nested loops where only the inner loop changes a variable

The original example still seems to not work.

Wrong optimization with nested loops where only the inner loop changes a variable

Something about nested loops is not going right, with the type analysis of `_re` there. Need to investigate.

Added optional `sub_collect` to `_detectImports`

Can you follow this to rebase: https://nuitka.net/doc/developer-manual.html#nuitka-git-github-workflow I want to take a look, what this changes, and what to take from it, and why.